Senior GPU Kernel Developer (AI)

TL;DR

Senior GPU Kernel Developer (AI): Implementing novel algorithms on GPUs and optimizing low-level kernels for high-performance AI applications with an accent on hardware limitations, profiling, and automation. Focus on building and refining complex computational kernels to improve efficiency and performance for an R&D-focused AI startup.

Location: Must be based in Poland (Remote or Hybrid at Warsaw office).

Company

An engineering-focused development firm specializing in R&D support for startups and disruptive technology companies.

What you will do

• Implement novel algorithms on GPUs, pushing hardware performance to its limits.
• Proactively develop and implement speedup and automation methods for GPU kernels.
• Build and operate low-level profiling setups to analyze and improve performance.
• Optimize and refine GPU kernels to achieve maximum efficiency.
• Collaborate closely with co-founders and technical leads on core technology development.
• Stay current with evolving hardware trends and new architecture capabilities.

Requirements

• Strong background in C/C++ and CUDA development.
• Extensive experience in code profiling and performance optimization using Nvidia tools.
• Deep understanding of GPU architecture and Nvidia hardware.
• Ability to work independently and collaboratively in a Polish-based or remote environment.

Nice to have

• Strong mathematical background (linear algebra, calculus).
• Deep understanding of AI, machine learning, and deep learning algorithms.
• MSc or PhD in Math, CS, EE, or a related field.
• Previous leadership or team lead experience.

Culture & Benefits

• Paid vacation: 20 business days per year.
• 100% sick leave payment plus 5 additional sick days.
• Medical insurance coverage after the probationary period.
• Equipment provision for your workspace.
• Partial compensation for professional educational costs and certifications.
• Legal and accounting support provided in Poland.
• Language classes in English and Polish twice a week.